On Wed, 25 Jun 2014, Richard Biener wrote:
On Wed, Jun 25, 2014 at 10:54 AM, Jakub Jelinek <ja...@redhat.com> wrote:
On Wed, Jun 25, 2014 at 10:14:17AM +0200, Richard Biener wrote:
Perhaps not unsurprisingly, the patch is very similar. Bootstrapped
and tested on x86_64-linux. OK for trunk?
Ok - please give the C++/atomics folks a chance to comment.
This change of default behavior should also be documented in
gcc-4.10/changes.html.
Do we want to allow the store data races by default with -Ofast even in strict
conformance modes (-std=c++11, -std=c++14, -std=c11)?
I think so. -Ofast means -Ofast (same issue with fp contraction and other
stuff -ffast-math enables that is not valid in strict conformance mode).
One thing I am missing is a -single-thread option that would allow races,
remove atomics and thread locals, etc, without breaking conformance as
long as no second thread is created. I hope that not too many libraries
use threads internally in a way that would break this.
--
Marc Glisse
